Did you check for longer time? We are still not sure where messages are lost (Gateway, Backend), but it is clear that looss rates are changing over the day. This are hourly packet losses over the last 7 days from a node, that sends one message per minute:
The gateway ist a kerlink iFemtocell, but we have similar issues on an ttn gateway (though total losses are lower). It is not yet clear what causes the losses, possibly an overload on the backend.